Discussion Paper No. 04-75

Capital Control, Debt Financing and Innovative Activity

Dirk Czarnitzki and Kornelius Kraft

Discussion Paper No. 04-75

Capital Control, Debt Financing and Innovative Activity

Dirk Czarnitzki and Kornelius Kraft

Die Discussion Papers dienen einer möglichst schnellen Verbreitung von neueren Forschungsarbeiten des ZEW. Die Beiträge liegen in alleiniger Verantwortung

der Autoren und stellen nicht notwendigerweise die Meinung des ZEW dar.

Discussion Papers are intended to make results of ZEW research promptly available to other economists in order to encourage discussion and suggestions for revisions. The authors are solely

responsible for the contents which do not necessarily represent the opinion of the ZEW.

Download this ZEW Discussion Paper from our ftp server:

ftp://ftp.zew.de/pub/zew-docs/dp/dp0475.pdf

Non-technical SummaryIt is generally accepted that technical progress, along with human capital, has amajor impact for the wealth and long-term growth of industrialized countries. SinceKnight, Schumpeter and the beginning of the discussion on innovation, greatemphasis has been placed on the central role of the entrepreneur in this process. Itseems to be generally accepted that the market is an entrepreneurial process, alearning process is central to the market, and entrepreneurial activities are creativeacts of discovery. In stark contrast to these statements on the entrepreneurs, largefirms are led by managers nowadays, which leads to a classical principal-agentproblem. It is questionable whether managers always act in the interest of the capitalowners. On the one hand, many firms may be owned by people or cooperations thathold a major share of the capital. In this case, one could argue that the managers ofthose firms are monitored by the shareholders and their major actions and decisionsare closely controlled. On the other hand, firms with widely held stock are usuallyassumed to be insufficiently controlled by the shareholders. Each shareholder facesthe well-known free-rider problem, as the benefits of an initiative monitoring themanager by one shareholder aimed at improving the quality of management will beenjoyed by all other shareholders. The consequence will be underinvestment intomonitoring the management. It remains to be investigated what consequences thishas for the innovation process.

First the paper discusses theoretically the incentive mechanisms driving innovationin managerial firms. There are opposing effects at work, and it is a priori unclearwhich incentive mechanism outweighs the other. On the one hand, managers mightbe reluctant to invest in innovation activity due to the inherent risk of failure. On theother hand, managers might invest above the profit-maximizing level into R&Dactivities, as it is a source of firm growth and the managers' salaries usually dependnot only on profits but on growth, too. In this line, debt financing is interesting tostudy. In the case where the positive growth incentive outweighs the risk argumentinfluencing innovation negatively, insufficiently controlled managers might bedisciplined by the credit market.

We then report the results of an empirical study on the effects of the concentration ofcapital shares and debt financing on innovative activity. For this purpose, a databaseon German manufacturing firms is used. In total we use information on 275 firms forthe years 1982 to 1996, which leads to 2,793 observations. Innovation is specified bythe number of patent applications per year.

It turns out that companies with widely held capital stock are more active ininnovation, i.e. weakly controlled managers show a higher innovation propensity.This supports the hypothesis that the positive growth incentive is dominating thenegative risk incentive. However, the higher the debt ratio, the more disciplined themanagers are.

Capital Control, Debt Financing and Innovative Activity

Dirk Czarnitzki* and Kornelius Kraft**

October 2004

Abstract

The present paper first discusses theoretically the different incentives ofmanager- versus owner-controlled firms for investment into innovativeactivity. In addition, the role of debt financing is analyzed. Subsequentlythe results from an empirical study on the determinants of innovativeactivity measured by patent applications are presented. A sample ofGerman firms covering 2,793 observations is used, and it turns out thatcompanies with widely held capital stock are more active in innovation,i.e. weakly controlled managers show a higher innovation propensity.However, the higher the leverage the more disciplined the managers are.

Keywords: Innovation, Patents, Corporate Governance, Limited Dependent VariablesJEL-Classification: C25, L11, O31, O32

* Centre for European Economic Research (ZEW)P.O.Box 10 34 4368034 MannheimGermanyPhone: +49 621 1235-158Fax: +49 621 1235-170E-mail: czarnitzki@zew.de

** University of DortmundVogelpothsweg 8744227 DortmundGermanyPhone: +49 231 755-3152Fax: +49 231 755-3155E-Mail: k.kraft@wiso.uni-dortmund.de

*, ** The authors are very grateful to the Deutsche Finanzdatenbank (DFDB) and to Prof. Dr. HansPeter Möller (Jahresabschlußdatenbank) for the very generous provision of their databases. We alsothank the Deutsche Forschungsgemeinschaft (DFG) for financial support. In addition ThorstenGliniars, Bianca Krol, Hartmut Meixner, Antonia Niederprüm, Stefan Schenk and Jörg Stank helped alot in data processing.This paper was partly written during a research stay of Dirk Czarnitzki at the University of Californiaat Berkeley. I thank the UCB for the hospitality and the Volkswagen Foundation for financial support.

1

1 Introduction

It is generally accepted that technical progress, along with human capital, has a majorimpact for the wealth and long-term growth of industrialized countries. Consequentlyeconomists have studied the causes and effects of innovation in theoretical andempirical studies for a long time.

Since Knight, Schumpeter and the beginning of the discussion on innovation, greatemphasis has been placed on the central role of the entrepreneur in this process. As anexample Kirzner (1985: 118), in line with the views of Hayek and Mises, emphasizesthat a) the market is an entrepreneurial process, b) a learning process is central to themarket and c) entrepreneurial activities are creative acts of discovery. In stark contrastto these statements on entrepreneurs, large firms are led by managers nowadays, andthis leads to a classical principal-agent problem. Managers are mainly paid on the basisof a fixed salary and it is rather questionable whether they always act in the interest ofthe capital owners. According to Mises (1951: 13) an entrepreneur is defined by thefollowing criterion: “There is a simple rule of thumb to tell entrepreneurs from non-entrepreneurs. The entrepreneurs are those on whom the incidence of losses on thecapital employed falls.” If this definition is used, the vast majority of the leading firmsin all developed capitalist countries are not led by entrepreneurs. It remains to beinvestigated what consequences this has for the innovation process.

The question of control by the shareholders is related to firm leadership by managers. Ifcapital is concentrated, the owners have clear incentives to monitor the managers, as -given the large amount of capital invested - the return from this activity is significant.Moreover, in most cases the shareholders are very familiar with the situation of a firmand its environment, which reduces the problem of asymmetric information. Thissituation is different with diversified shares, as it is frequently the case in the moderncorporation. With diversified shares the well-known free-rider problem arises, becauseevery activity of an individual shareholder in order to increase efficiency of themanagement has to be shared with numerous other shareholders. Furthermore, theindividual shareholder with widely diversified shares has neither the knowledge of thespecific situation the firm is facing nor does she/he have a monetary incentive tointervene, because the financial stake is only small.

Recently some theoretical contributions have elaborated on this issue. Holmström(1989) discusses the principal-agent problem in the context of innovation. Zwiebel(1995) uses a two-period model with the expected remuneration of managers andconsiders the incentives for innovation if a manager is dismissed, provided theperformance is below some level. Aghion et al. (1997) apply an intertemporal model todemonstrate that managers will invest less in innovation than others. They assume thatthe managers have private costs connected with innovation, and these determine theirresults. According to Zwiebel (1995) as well as Aghion et al. (1997) the manager-led

2

firm is less innovative than the owner-led company. Both articles are based on risk-neutrality on the part of the managers, which has no consequences in manycircumstances. However, as will be discussed below, the attitude concerning riskdetermines the profit-orientation of the managerial remuneration in this context, andmonetary incentives in turn affect the efforts of the managers.

Another strand of literature discusses possibilities to overcome the incentive problemsin the principal-agency situation of the managerial firm. One possible instrument is thecapital structure. It is argued (among others by Dewatripont and Tirole, 1994, Hart andMoore, 1995) that debt keeps a tight reign on managerial agency costs. The governancerole of debt comes from the threat of bankruptcy (Aghion and Bolton, 1992), thereduction of free cash flow (Jensen, 1986) and monitoring by the creditors (Diamond,1984).

There are numerous empirical papers on incentives for managers, such as thedetermination of remuneration and the behavior of the managerial firm with respect togrowth, investment, profitability and other topics. However the relation betweencorporate governance and innovative activity is much less analyzed. The only empiricalresearch on this issue, as far as we are aware, is our own (Czarnitzki and Kraft,2004a,b).

The purpose of the present article is to report the results of an empirical study on theeffects of the concentration of the capital shares and debt financing on innovativeactivity. For this purpose, a database on German manufacturing firms is used. In totalwe use information on 275 firms for the years 1982 to 1996, which leads to 2,793observations. Innovation is specified by the number of patent applications per year. Amajor difference in comparison to earlier research in this field is the consideration of thecapital structure.

2 Theory

The separation of ownership and control leads to a fundamental control and incentiveproblem. In the presence of asymmetric information it is a not a trivial task to design anincentive mechanism, where the agent (manager) works in the interest of the principal(shareholder). The problem has several dimensions, as it not only matters what themanager does, but also how much effort is put into reaching a certain task. Therefore,efficient effort levels are required for optimal allocation of resources. One simplesolution would be control by the outside investors. Clearly this may be a possibility ifthere is a single dominant investor; however, this is rather unrealistic with widelydiversified capital shares. Each shareholder faces the well-known free-rider problem, asthe benefits of an initiative of one shareholder to improve the quality of managementwill be enjoyed by all other shareholders. The consequence will be underinvestmentinto monitoring the management and improving its decisions. “As a result, the

3

management of a company with many shareholders will be under little pressure toperform well” (Hart, 2001).

Investment into R&D is clearly a risky project. In fact one of the most important taskswhen innovative activity is analyzed is the consideration of the effect of risk. In thegiven context, risk has essentially two different effects: First, it affects the behavior ofmanagers if they can control risk. Second, it determines the relation of fixed versusvariable (profit-related) compensation for exogenously given risk.

Incentives and Risk

Starting with the second point, the incentive problem of managers has been analyzed innumerous articles. In the standard principal-agent problem as employed by Holmströmand Milgrom (1987), pay-performance sensitivity depends on the parameters of theproduction function that transforms executive effort into output, and the distribution ofa random variable which affects corporate performance. Pay-performance sensitivity isalso influenced by the form of the manager’s utility function, especially the degree ofher/his absolute risk aversion1 and her/his cost of effort which is taken to be strictlyconvex. Holmström and Milgrom (1987) show that it is not ideal to make theremuneration totally output-dependent, but to use a combination of a fixed and aflexible share. The variable part of remuneration is a decreasing function of riskaversion. A contract which only depends on output without fixed remuneration wouldnot be accepted by risk-averse managers. Thus, the optimal contract is a compromisebetween a partial insurance of the agent and the provision of incentives.2

The Holmström-Milgrom model has two limitations for the analysis of the presentproblem: dismissals are neglected and risk is exogenously given. Holmström andMilgrom do not consider the possibility of dismissals. However, if the manager does notmeet a certain standard, in many cases she/he will be dismissed (for empirical evidenceon this issue see Weisbach, 1988, Warner et al., 1988, Gilson, 1989, as well as Fizel andLouie, 1990). Hence the consideration of risk and risk aversion is even more complexthan in the otherwise very useful model of Holmström and Milgrom (1987). If a

1 The existing theoretical contributions on the relation of management-leadership and innovative activitylike Zwiebel (1995) and Aghion et al. (1997) neglect risk aversion. 2 There are many extensions and alternatives to this model: Holmström and Milgrom (1991) prove that, inmultitask environments, fixed payments are useful because of the unobservability of some importanttasks. Baker (1992) as well as Baker et al. (1994) consider the situation where the principal's objectivecannot explicitly be the subject of an employment contract due to some vagueness in the objectivesthemselves. Given this vagueness it is quite likely that in such a situation a performance measureresponds to the agent’s actions in a different way than the principal’s objective responds to these actions.If the objective is difficult to define, the firm will reduce the sensitivity of the incentives to theperformance measure. The contract will be inefficient even if the agent is risk-neutral. Laffont and Tirole(1986, 1990, 1993) develop a model on the incentive payments to regulated monopolists, which has beenapplied by Horn et al. (1994) to the theory of management incentives. Additional explanations of thelimited application of profit-dependent payments are given by Perri (1994) and Palley (1997).

4

manager pursues a risky project like R&D, for example, a failure is definitely possible,and this would have a negative impact on profits. Given that there is some relationbetween profitability and managerial compensation, the manager’s income woulddecrease as well. If the R&D project is costly, the person might even be dismissed.Perhaps the Holmström-Milgrom model could be enlarged by the consideration ofdismissals as a strong penalty, but based on the assumption that this possibility has anexogenously given probability.

This theory of income determination (possibly enlarged on the dismissal threat) mayexplain the variable and fixed shares of compensation. This will in turn determine thebehavior of the agents, in particular, efforts react to flexible rewards. If an efficientR&D management implies (quite realistically) effort on the part of the manager with aprofit-share below one hundred percent, the manager will not exert the same effort levelas the entrepreneur would do. As a consequence, either less R&D will be carried out orthe same R&D expenditures will be allocated less efficiently.

The second effect of risk is even more pronounced. In contrast to what is assumed in theHolmström-Milgrom model, the manager is able to control the level of risk byallocating firm resources to more or less risky investments, when he or she tries toexpand firm activities (in many cases above the profit-maximizing level). It is mostlikely that innovative activity is more risky than for example increasing marketingexpenditures for existing products, raising the production capacity without altering theproduction process itself, or mergers and acquisitions. Hence the risk-averse managerwill prefer less risky investments to risky projects, which is an argument against R&Dactivities.

Things become even more complicated if the standard assumption of a risk-neutralentrepreneur is put into question. It is reasonable to assume risk-neutrality by the capitalsuppliers (investors) if these diversify their fortune over many different firms and thuseffectively minimize risk. Such atomistic shareholders ask for a maximization of theexpected value of investment irrespective of risk considerations. In contrast to atomisticshareholders, the classical entrepreneur very probably holds an undiversified portfolioand has concentrated his/her wealth in his/her firm. The owner therefore has goodreasons to be risk-averse. If this scenario is true, the stock company with broadlydiversified shares should be efficient and risk-neutral while the owner-led firm wouldbe risk-averse.

Growth incentives

Whilst the consideration of risk is not trivial, there is an important second factorinfluencing R&D activities, i.e. growth. Numerous studies have demonstrated thatmanagers’ salaries largely depend on the size of the firm in question and to a muchsmaller degree on profitability. For example, Baker et al. (1988) cite evidence that, as ageneral rule, the salary of top executives increases by 3% when the firm’s sales increaseby 10%. In an examination of US companies, Jensen and Murphy (1990) found that a

5

one dollar increase in the total wealth of capital owners (stock value and dividend) leadsto only a three cent increase in management’s total compensation.3 This will clearlylead to a growth orientation of managerial firms. Non-monetary determinants ofmanagers’ utility, such as status, power and prestige, are probably also related moreclosely to firm size than to profitability. Perhaps the capital market realizes the financialsuccess, but public opinion is more impressed by employment figures and thus thepersonal influence increases with firm size. In a dynamic setting, the manager’s incomedepends on firm growth, and since the beginning of this whole strand of literature,economists have hypothesized that the managerial firm is more growth-orientated thanthe owner-led company. Growth is, among other factors, determined by innovativeactivity (cf. e.g. Fagerberg, 1987).

Another argument supports the preferred use of innovative activity as an instrumentgenerating growth. Public opinion concerning the expansion activities of managerial-ledfirms have recently become quite critical. However, this is mostly directed towardsmergers and acquisitions and capacity expansions of established product lines. Amongstthe public, investments into R&D are usually quite favorably evaluated, as they areexpected to raise the competitiveness of the firm, and, in aggregate, that of the wholeeconomy. Moreover, it increases growth possibilities and therefore income in the longrun. Even politicians frequently use figures like R&D and patent statistics as “successnews” without questioning whether this is really an efficient allocation of resources. Insuch a situation, the management will face less objections when pursuing R&D projectsinstead of other growth strategies. These arguments imply higher R&D expendituresand more innovations by the managerial firm. Therefore, we have opposing incentivesfor the managers and it is an empirical question as to which are the predominant ones.

Debt Financing

A suggestion for solving the agency problem (or for reducing it) is by makingcommitments. As discussed above, profit-related pay is of limited use in the presence ofrisk-averse managers. An alternative commitment instrument is debt, because debtpayments are fixed ex ante and cannot be circumvented. Therefore, debt commitmentsare “harder” than dividends and are able to control management in situations ofasymmetric information (see Jensen, 1986, Hart and Moore, 1995, Harris and Raviv,1990, Stulz, 1990, Zwiebel, 1996). If the management is unable to meet debt payments,the maximum penalty is bankruptcy. So this financial policy is effectively designedtowards curtailing overspending on the part of the management, as this behaviorincreases the likelihood that pay-out commitments will be met. It can be ensured that

3 Many studies besides Jensen and Murphy (1990) have examined the dependence of managementcompensation on performance and found a limited relation (c.f. Gibbons and Murphy, 1990, Main, 1991,Kaplan, 1994a,b, Kraft and Niederprüm, 1999). In contrast, Hall and Liebman (1998) find quite a strongrelationship between firm performance and managers’ remuneration.

6

internal funds are allocated to profitable investments instead of to personal aims of thetop managers.

Another argument in favor of debt financing is the relative cost advantage compared toretained earnings and the issuing of new shares (cf. Hall, 2002). In addition, debt mightbe regarded as a positively valued signal about the true value of the firm in a situationof asymmetric information (see Ross, 1977). According to these views, increasingleverage has only beneficial effects. While debt is usually advantageous if the prospectsare good, counter-arguments exist, too: if a firm faces a recession, the bankruptcy riskincreases with its degree of leverage. This determines the disciplinary effect fromabove, and, hence, an optimal degree of leverage is sometimes postulated.

Since the work of Schumpeter it is assumed that debt itself (without consideration ofleadership and incentive problems) has a negative correlation with innovativeness. First,R&D cannot be used as a collateral in credit negotiations with banks. In contrast to aninvestment in physical assets that is capitalized in the firm's balance sheet, R&D is anexpense and thus recognized as sunk cost at the time it is spent. Second, due to theuncertainty with respect to the outcome of an R&D project, an asymmetric informationproblem between borrower and lender emerges. As a result, banks and other possibleinvestors are reluctant to finance such investments (see Hall, 2002).

3 Empirical set-up and data

The basis for the empirical investigation is a random sample of Western German joint-stock companies. The main part of the database comprises of annual accounts collectedby the Deutsche Finanzdatenbank. The number of firms observed has been extended bysampling firms from the Hoppenstedt "Bilanzdatenbank" and from annual reports.Information on the governance structure has been added from various sources (mainlythe "wer gehört zu wem" publications of the Commerzbank). Furthermore, the numberof patent applications per year have been linked to the firm level data. In total, thesample consists of 275 manufacturing firms observed in the period from 1982 to 1996yielding an unbalanced panel of 2,973 firm-year observations. This database is enrichedby information on the industry level in order to test for the impact of national orinternational competition.

The theory presented above deals with the difference between the classicalentrepreneurial firm and the modern managerial firms. This question is tackled in thefollowing way: firms with widely held stock are compared to those controlled by adominant capital owner. We identify a dominant capital owner as a person or firmholding at least 25% of the capital shares. If the dominant capital owner is a family(which is not a rare case in Germany), we employ the joint capital share as the relevantyardstick. A dummy variable DISPERSION indicates whether the capital stock of acompany is widely held (DISPERSION=1) or if a dominant owner exists(DISPERSION=0). This dummy variable takes unit value in about 19% of all cases. In

7

line with the literature, we suppose that in firms with widely held stock(DISPERSION=1), management is insufficiently controlled.

While no information on R&D is available, patent counts represent an alternativemeasure of innovation. Although a patent measure differs from R&D4, both variablesare highly correlated, because patents can only reasonably be filed (and granted) ifcorresponding R&D efforts have been undertaken. Therefore, it appears justified to usepatents as an indicator describing innovative activity. Patents may also have anadvantage over R&D: it is quite likely that large (managerial) firms conduct refinementson existing products and processes and therefore perform R&D, but they do not investinto research on more radical inventions (see e.g. Henderson, 1993), which might bebetter qualified for filing patent applications. Thus patents might even reflect the reallyrisky R&D investments better than mere R&D expenditure, which could lead to a bettertest of the risk argument in managers' incentive schemes. Clearly patents have alsodisadvantages as an innovation indicator, because some inventions are never patentedand some patented knowledge is never implemented into new products or processes.The study analyzes the innovative activity of firms specified as the number of patentapplications filed per year. This is a standard variable in studies on technical progressand is henceforth called PATENTS.

An important new feature of the present study compared to earlier ones (Czarnitzki andKraft, 2004a, b) – in addition to the new data source – is the inclusion of the firms'financial structure. We consider the financial situation by the ratio of debt to totalcapital (DEBT). Debt has two potential effects in our empirical model: on the one hand,it is frequently argued that innovative activity has to be financed by internal resourceslike cash flow. The reasons for this are that bank managers are usually unable toevaluate the potential success of an R&D project and prefer to use physical assets tosecure loans (see Hall, 1992). If a firm does not achieve sufficient cash flow and is inneed of external financial support like debt, such a company will most likely show lessinnovative activity than others. Thus, the debt ratio is used as an indicator for externalfinancing, and we expect a negative relation to innovative activity. On the other hand,as discussed above, management might be disciplined by debt commitment. Therefore,we introduce the interaction variable DISPERSION*DEBT in order to test for thishypothesis. The expected coefficient of this variable depends on the sign of thecoefficient of DISPERSION: in the case of a positive coefficient of DISPERSION,firms with widely held stock would file more patents than others. We would concludethat the management does too much in this respect, at least if profit maximization isconcerned. The interaction variable is then expected to have the opposite coefficient asthe management is disciplined by the presence of debt commitment and the threat of apossible bankruptcy. The situation is a bit more complicated if the managers do not

4 While R&D is an input of the innovative process, patents represent an (intermediate) output.

8

conduct enough R&D, as they want to avoid the risk and do not care about the long-runconsequences for the firms’ values (negative coefficient of DISPERSION). In principle,capital market control is expected to lead to an improved, if not optimal, situation.Commitment by debt implies a reduction of free cash flow and therefore thepossibilities for any engagement above the absolutely necessary level are reduced. Inthe case where firms with dispersed ownership already show less innovative activity, itis however unlikely that DEBT*DISPERSION reduces innovativeness even further.The interaction variable would most likely have an insignificant coefficient in such acase. Less plausible, but not excludable, is the situation where external financing is usedto increase the competitiveness of a firm in order to enable survival. New financialresources would be used for the introduction of new products and an improvement ofthe technology used. In this scenario, a positive coefficient of DEBT interacted withDISPERSION is possible. The management knows that the new resources have to beused efficiently as the probability of bankruptcy increases significantly otherwise.5 Thisargumentation is, however, in contradiction to the discussion on the general difficultiesof financing innovation activities by external capital and therefore we find thisargumentation less convincing.6

We also experimented with Cash-Flow as the internal source of financing innovationactivities. It, however, turned out to be insignificant in all specification that have beenconsidered. Therfore, we dropped it from the final estimations presented below.

Of course, a number of other control variables have to be taken into account. As thenumber of patent applications will be correlated with firm size, and firm size isprobably related to the dispersion of capital shares, it is of primary importance tocontrol for size effects. Here, the number of employees (EMP) and its squared value areused.7

Three variables control for heterogeneous national and international competitivepressure: the share of sales exported (EXPORT) is measured on the individual firmlevel and describes the participation in international competition. A related variable issales of foreign firms in the domestic market compared to total sales of both foreign anddomestic firms (IMPORT). The import quota is measured at the industry level andexpresses competitive pressure from other countries, which is highly significant for anopen economy like Germany. Another variable controlling for domestic competition is

5 It is beyond the scope of this paper to analyze the question, why some firms show a higher debt ratiothan others. In our framework debt is predetermined, which is reflected in the use of DEBT in t-1.6 The analysis becomes even more complicated if the decision rights in different economic situations areconsidered. According to Aghion and Bolton (1992) the external investor should obtain full control if afirm performs poorly. If the firm performs well, the entrepreneur should have the decision rights. See alsoKaplan and Stromberg (2003) for empirical evidence on these predictions. 7 We also checked different specifications like log(EMP) or a number of size dummies, but the results donot depend on the particular specification of the size variables.

9

seller concentration (CR6), which is defined as the sales share of the six largestcompanies in an industry. If a firm in our sample is active in more than a singleindustry, we assign the concentration index from that industry where the largest share oftotal sales volume is realized. We assume that the consideration of international tradeand the concentration index allow to identify the degree of competitive pressure a firmfaces.

Capital intensity (KAPINT) is defined as the total capital stock divided by the numberof employees and indicates the technology intensity of firms' production. It issometimes interpreted a proxy variable for barriers to entry. Technology intensity isclearly an argument in favor of innovativeness, while the role of barriers to entry andtherefore competitive pressure is unclear and controversially discussed.

It is possible that younger firms are more innovative than more established firms, as thestart-up phase usually goes hand in hand with the commercialization of inventions. Asdescribed above, the established firms might be more reluctant to introduce"fundamental" innovations. We take this possible relationship into account by includingthe log of firms' age in the analysis (ln(AGE)).

The opportunities for innovation will certainly differ among industries. Therefore 12industry dummies are included in all regressions. Furthermore, we add 15 timedummies in order to control for changes in the general attitude to patent newly createdknowledge and macroeconomic influences. Descriptive statistics of the variables usedcan be found in Table 1. All exogenous variables are lagged once, in order to treat thecovariates as predetermined, and thus avoid endogeneity problems.

Table 1: Descriptive statisticsDISPERSION = 0

N0 = 2,340DISPERSION = 1

N1 = 543Variable Mean Std. Dev. Min. Max. Mean Std. Dev. Min. Max.PATENTit 13.61 43.90 0 1088.00 87.30 247.67 0 1702.00(PATENT/EMP/1,000)it 3.35 7.45 0 105.37 5.48 9.08 0 76.87(EMP/1,000)i,t-1 3.44 8.34 0.01 61.62 8.67 17.06 0.05 71.81ln(AGE)i,t-1 4.52 0.72 0 6.53 4.64 0.43 1.95 5.86EXPORTi,t-1 30.99 22.71 0 94.90 37.32 23.41 0 97.65KAPINTi,t-1 0.22 0.43 0.00 9.75 0.46 1.22 0.03 10.04DEBTi,t-1 0.39 0.19 0.03 0.97 0.36 0.19 0.02 0.85CR6i,t-1 23.94 18.43 6.50 88.00 23.79 16.72 6.50 88.00IMPORTi,t-1 0.29 0.20 0.01 1.36 0.35 0.23 0.03 0.99Note: 12 industry dummies and 15 time dummies not presented.

As Table 1 shows, the firms with widely dispersed stock are much larger: on average,such firms have about 8.7 thousand employees, whereas the sample mean for firmshaving a dominant capital owner is only 3.4 thousand. This leads, of course, to a largeaverage difference in the number of patent applications filed per year: 87.3 applicationsfor firms with DISPERSION=1, and only 13.6 for firms with dominant capital owners.We account for the correlation of DISPERSION and EMP in the upcoming analysis by

10

estimating regressions not only on PATENT, but also on PATENT/EMP which can beinterpreted as a kind of an innovation intensity. Rescaling PATENT by the number ofemployees reduces the size effects in the variable. On average, the firms controlled by adominant capital owner show about 3.4 patent applications per thousand employees,whilst this figure is 5.48 for the other firms. Other interesting differences are that firmswith widely dispersed stock realize higher export quotas and show a higher capitalintensity, on average.

4 Econometric Results

As mentioned above, we consider PATENT as dependent variable, and run count dataregressions; in particular Negative Binomial models (NegBin).8 Count data models arebased on the assumption that the exogenous variables describe the move from 0 to 1 inmuch the same way as, for example, from 18 to 19. However, many firms do noresearch at all and therefore they will never file a patent. Hence, in many cases there is afundamental difference between observations with the value zero and those which haveone or more patents. For this reason, the so-called zero-inflated models have beendeveloped. The first equation analyzes whether the dependent variable is zero or largerthan zero. Any observation larger than zero is adjusted to one, and then either a Logit ora Probit model is applied. In the second equation, the NegBin model is applied for thoseobservations where the dependent variable is larger than zero (see e.g. Cameron andTrivedi, 1990, for details of econometric count data models). The share of censoredobservations in the full sample amounts to almost 49%. In fact, the sample containsfirms that never file a patent application during the whole period observed. Therefore,we also consider the subsample of patenting firms in the second stage. The subsampleof firms that file at least one patent between 1982 to 1996 consists of 1,950 firm-yearobservations and the share of censored observations is less than 27%.

Due to the fact that DISPERSION is correlated with EMP, we also estimate regressionsusing PATENT/EMP as dependent variable, in order to control for possible size effects.For this, we consider OLS estimations and Tobit regressions in order to account for theleft-hand censoring of the dependent variable's distribution. One serious problempossibly occurring in Tobit estimations is heteroscedasticity, because a violation of theassumption of homoscedasticity leads not only to biased standard errors, but also toinconsistent coefficient estimates. Thus we consider multiplicative groupwiseheteroscedasticity and replace the homoscedastic variance s by ( )'expi iwσ σ α= ,

where wi represents the regressors entering the heteroscedasticity term and a theadditional coefficients to be estimated. The regressors in the Tobit model's

8 The Poisson model is based on the assumption of equality of the conditional mean and varianceVar(y|x)=E(y|x). This restrictive assumption is released in the Negative Binomial model.

11

heteroscedasticity term are seven firm size dummies (based on the number ofemployees), twelve industry dummies and fourteen time dummies.

Although our database would, in principle, allow controlling for firm-individual fixedeffects in the regressions, we cannot unfortunately run panel models. Our main variableof interest, DISPERSION, does not change over time (except in a few very rare cases),i.e. we cannot identify the coefficient of DISPERSION in fixed effects regressions.Therefore, we rely on pooled cross-sectional estimations in this paper.

Table 2 presents the estimation results using the full sample. Before discussing theresults, we will report some test statistics. We tested for heteroscedasticity in the Tobitmodel by computing LM tests (see e.g. Greene, 2000) using 7 size dummies[LM((c2(7)) = 179.27)], 12 industry dummies [LM((c2(12)) = 203.76)] and 14 timedummies [LM((c2(14)) = 139.76)]. The test statistics indicate that homoscedasticity isrejected in all cases. Thus the coefficients in the Tobit model might be misleading dueto the violation of the homoscedasticity, and instead the results of the heteroscedasticTobit model are preferable.

In the count data model, we tested for overdispersion and the LR test indicated that thePoisson model (assuming no overdispersion) is clearly rejected [LR(c2(1)) = 43,464.41].In addition, we computed a Vuong test in order to check if the zero-inflated negativebinomial (ZINB) outperforms the negative binomial regression. We include ln(EMP),ln(AGE), EXPORT, IMPORT, CR6, DEBT, KAPINT in the inflation equation, and findthat the Vuong statistic is significant at the 1% level [VUONG (c2(1)) = 8.72].

The main results are very robust over all different estimated models. The coefficient ofDISPERSION is positively significant in all cases. It turns out that the positive growthincentives to innovate outweigh the negative risk incentives. Thus managers that areinsufficiently controlled possibly spend too many resources on innovation compared tothe profit-maximizing level.

Furthermore, DEBT shows a negative coefficient in the regressions (except in the zero-inflated negative binomial model). This supports the hypothesis that firms facefinancing constraints for R&D. A very interesting result is that firms with widely heldstock innovate more than those firms with a dominant capital owner. However, as thecoefficient estimates DISPERSION*DEBT are negatively significant in all regressions,we can conclude that insufficiently shareholder-controlled managers are disciplined bythe credit market.

12

Table 2: Estimation results with full sampleDependent variable: (PATENT/EMP/1,000)it PATENTit

Variable OLS Tobit Tobit with hetero-scedasticitya) NegBin ZINBb)

(EMP/1,000)i,t-1 0.265 *** 0.720 *** 0.344 *** 0.351 *** 0.257 ***(0.053) (0.084) (0.050) (0.014) (0.012)

(EMP/10,000)2i,t-1 -0.426 *** -1.120 *** -0.502 *** -0.470 *** -0.335 ***

(0.091) (0.142) (0.080) (0.024) (0.019)ln(AGE)i,t-1 0.115 0.126 -0.036 -0.059 -0.084 *

(0.172) (0.374) (0.226) (0.046) (0.045)EXPORTi,t-1 0.023 ** 0.046 *** 0.039 *** 0.014 *** 0.016 ***

(0.010) (0.013) (0.009) (0.002) (0.002)KAPINTi,t-1 1.701 *** 2.181 *** 1.779 *** 0.161 *** 0.189 ***

(0.651) (0.365) (0.397) (0.057) (0.056)DEBTi,t-1 -1.403 ** -6.416 *** -3.999 *** -0.925 *** -0.343

(0.691) (1.572) (0.928) (0.233) (0.234)CR6i,t-1 -0.049 *** -0.095 *** -0.034 * -0.014 *** -0.005 *

(0.017) (0.026) (0.018) (0.003) (0.003)IMPORTi,t-1 1.313 -0.013 0.842 0.146 0.638 **

(0.890) (2.158) (1.167) (0.316) (0.287)DISPERSIONi,t-1 4.626 *** 5.327 *** 3.178 *** 0.520 *** 0.470 ***

(1.066) (1.418) (0.919) (0.184) (0.154)DISPERSION*DEBTi,t-1 -8.187 *** -8.710 ** -6.820 *** -1.030 ** -1.100 ***

(2.558) (3.560) (2.372) (0.485) (0.420)Intercept 1.016 -2.657 0.286 0.171 0.612 *

(1.378) (2.981) (1.962) (0.352) (0.369)Industry dummies F(14, 2756)

= 22.00***c2(12)

= 558.31***c2(12)

= 149.62***c2(12)

= 580.26***c2(12)

= 588.87***Time dummies F(12, 2756)

= 2.41***c2(14)

= 24.67**c2(14)= 15.97

c2(14)= 30.28***

c2(14)= 37.29***

# of observations 2,793 2,793 2,793 2,793 2,793Log likelihood - -5,900.02 -5,552.92 -6,580.48 -6,411.55R2

/ McFadden R2 0.190 / - - / 0.095 - / 0.148 - / 0.178 - / 0.199Standard errors in parentheses. *** (**,*) indicate a significance level of 1% (5, 10%).a) The heteroscedasticity term includes 7 size dummies, 12 industry dummies and 14 time dummies(coefficient estimates not reported).b) The inflation equation includes ln(EMP), ln(AGE), EXPORT, IMPORT, CR6, DEBT, KAPINT(coefficient estimates not reported).

The other results are interesting as well: there is a strong size dependence of patentingactivity. All models report a significant inversely u-shaped relationship. Moreover,competition affects innovative activity. The more firms are engaged on internationalmarkets (as measured by EXPORT) the more they rely on innovation activities. Incontrast, the lower domestic competition firms face (as measured by CR6) the less theytend to innovate. In addition, the higher the capital-intensity the firms' production(KAPINT), the more they innovate. We do not find age effects, possibly due to highcorrelation with firm size. Finally, there are significant differences in innovating overindustries and time (see joint hypothesis tests reported for industry and time dummies).

13

As a further robustness check of our results we drop the firms that never filed any patentin the observed period. The sample size reduces to 1,950 observations using thissubsample. Note that we still estimate Tobit models and zero-inflated negative binomialmodels, because even the patenting firms do not patent in every period. Therefore, westill want to take left-censoring of the distribution of the dependent variables intoaccount.

The econometric test statistics yield similar results as reported for the full sample. Theestimation results for using the subsample of patenting firms are reported in Table 3. Ingeneral, the estimated effects of the covariates are not as pronounced as in the fullsample, but the main results on capital control hold in the subsample, too. There theestimated effects even become slightly stronger.

Table 3: Estimation results with subsample of patenting firmsDependent variable: (PATENT/EMP/1,000)it PATENTit

Variable OLS Tobit Tobit with hetero-scedasticitya) NegBin ZINBb)

(EMP/1,000)i,t-1 0.129 ** 0.334 *** 0.147 *** 0.286 *** 0.258 ***(0.059) (0.083) (0.041) (0.012) (0.014)

(EMP/10,000)2i,t-1 -0.236 ** -0.539 *** -0.225 *** -0.374 *** -0.334 ***

(0.097) (0.140) (0.063) (0.020) (0.021)ln(AGE)i,t-1 -0.001 -0.016 0.094 -0.076 * -0.066

(0.232) (0.380) (0.194) (0.043) (0.052)EXPORTi,t-1 0.027 ** 0.039 *** 0.028 *** 0.015 *** 0.014 ***

(0.014) (0.013) (0.009) (0.002) (0.003)KAPINTi,t-1 2.333 *** 2.489 *** 2.056 *** 0.150 *** 0.132 **

(0.806) (0.375) (0.371) (0.054) (0.053)DEBTi,t-1 0.337 -1.410 -2.423 *** -0.298 -0.302

(1.082) (1.654) (0.880) (0.216) (0.225)CR6i,t-1 -0.048 ** -0.055 ** -0.002 -0.008 ** -0.006 **

(0.020) (0.026) (0.015) (0.003) (0.003)IMPORTi,t-1 2.281 * 4.383 * 3.377 0.717 *** 0.679 **

(1.249) (2.269) (1.076) (0.278) (0.277)DISPERSIONi,t-1 5.700 *** 6.090 *** 2.692 *** 0.518 *** 0.496 ***

(1.333) (1.420) (0.909) (0.163) (0.164)DISPERSION* DEBTi,t-1 -10.804 *** -12.083 *** -7.157 *** -1.183 *** -1.163 ***

(3.531) (3.633) (2.291) (0.445) (0.438)Intercept 2.359 -0.079 0.514 0.424 0.518

(2.017) (3.132) (1.770) (0.350) (0.391)Industry dummies F(12, 1913)

= 19.46***c2(12)

= 261.52***c2(12)

= 234.85***c2(12)

= 425.37***c2(12)

= 457.53***Time dummies F(14, 1913)

= 2.94***c2(14)

= 30.60***c2(14)

= 17.39***c2(14)

= 32.51*** c2(14)

= 37.26***# of observations 1,950 1,950 1,950 1,950 1,950Log likelihood - -5,622.28 -5,184.31 -6,174.25 -6,132.98R2 / McFadden R2 0.157 / - - / 0.044 - / 0.118 - / 0.151 - / 0.156Standard errors in parentheses. *** (**,*) indicate a significance level of 1% (5, 10%).a) The heteroscedasticity term includes 7 size dummies, 12 industry dummies and 14 time dummies(coefficient estimates not reported).b) The inflation equation includes ln(EMP), ln(AGE), EXPORT, IMPORT, CR6, DEBT, KAPINT(coefficient estimates not reported).

14

Again, firms with widely held stock are more engaged in innovation activities, as thecoefficient of DISPERSION is significantly positive in all regressions. Managers ofthose firms are disciplined by credit markets, because the interaction termDISPERSION*DEBT is significantly negative in all models. However, DEBT is nolonger significantly different from zero in this setting. It is likely that possible financialconstraints deter firms from innovation completely, because they are not able to raisesufficient external capital. Looking only at innovating firms, the role of financialconstraints might be underestimated. Similar results are also reported by Czarnitzki(2002) who studies the impact of financial constraints on R&D expenditure in a sampleof German small and medium-sized manufacturing firms. The hypothesis that it isimportant to take the innovation status into account is also supported in the ZINBmodel. While DEBT is not significant in the continuous part of the patent distribution(PATENT>0), DEBT is significant in the inflation equation (coefficient estimate notreported in Table 3).

5 Conclusion

This paper considers the impact of management leadership of a firm in the presence ofweak capital market control on the incentives for innovative activity. There areopposing effects at work, as managers will dislike the risk associated with R&Dprojects but will like the growth effects of successful innovations. Hence the net effectof managerial leadership on innovative activity is unclear. Furthermore, we discuss therole of debt in R&D financing in general and in disciplining the management.

Subsequently the results from an empirical study are reported. We investigate the effectof a dispersed ownership of a firm compared to concentrated capital shares on thenumber of patents. The firms where the capital shares are broadly distributed file ceterisparibus more patents than the other companies.

Moreover, the role of debt financing is analyzed. Debt financing itself has a negativeimpact on innovativeness. This result is of interest, but most probably not verysurprising. However, debt has another effect: it disciplines the management of firmswith widely distributed shares and leads to behavior in line with firms which have adominant capital owner. Hence debt really has the effect of reducing the agencyproblem suggested in theoretical literature. To the best of our knowledge, this is the firstempirical test concerning the effect of debt financing on the behavior of management-led firms.

The implications of our empirical results appear to be quite interesting: managementleadership does not only have negative effects. Firms led by managers have morepatents than comparable companies. As patents are in a way the output of innovativeactivity (of an input like R&D) managements’ efforts to increase innovativeness cannotbe a pure waste of resources. However this innovativeness might be above the profit-maximizing level, as the management puts more weight on growth and less on the rate

15

of return on capital. As innovation usually has positive spill-overs it is frequentlyargued that the profit-maximizing firm does not carry out enough R&D. Hence in thiscase perhaps the managerial firm is closer to the socially efficient level ofinnovativeness than the traditional capitalist firm, but of course not because themanagement wants to maximize social welfare.

References

Aghion, P. and P. Bolton, (1992), An Incomplete Contracts Approach to Financial Contracting,Review of Economic Studies 59, 473-94.

Aghion, P., M. Dewatripont and P. Rey (1997), Corporate Governance, Competition Policy andIndustrial Policy, European Economic Review 41, 797-805.

Baker, G.P. (1992), Incentive Contracts and Performance Measurement, Journal of PoliticalEconomy 100, 598-614.

Baker, G.P., M.C. Jensen and K.J. Murphy (1988), Compensation and Incentives: Practiceversus Theory, Journal of Finance 43, 593-616.

Baker, G.P., Gibbons, R. and K.J. Murphy (1994), Subjective Performance Measures in OptimalIncentive Contracts, Quarterly Journal of Economics 109, 1125-1156.

Cameron, A.C. and P.K. Trivedi (1998), Regression Analysis of Count Data, Cambridge.

Czarnitzki, D. (2002), Research and Development: Financial Constraints and the Role of PublicFunding for Small and Medium-sized Enterprises, ZEW Discussion Paper No. 02-74,Mannheim.

Czarnitzki, D. and K. Kraft (2004a), Management Control and Innovative Activity, Review ofIndustrial Organization 24(1), 1-24.

Czarnitzki, D. and K. Kraft (2004b), Firm Leadership and Innovative Performance: Evidencefrom Seven EU Countries, Small Business Economics 22(5), 325-332.

Dewatripont, M. and J. Tirole (1994), A Theory of Debt and Equity: Diversity of Securities andManager-Shareholder Congruence, Quarterly Journal of Economics 109 ,1027-1054.

Diamond, D.W. (1984), Financial Intermediation and Delegated Monitoring, Review ofEconomic Studies 51, 393-414.

Fagerberg, J. (1987), A Technology Gap Approach to Why Growth Rates Differ, ResearchPolicy 16, 87-99.

Fizel, J.L. and K. Louie (1990), CEO Retention, Firm Performance and Corporate Governance,Managerial and Decision Economics 11, 167-176.

Gibbons, R. and K. Murphy (1990), Relative Performance Evaluation for Chief ExecutiveOfficers, Industrial and Labor Relations Review 43 (supplement), 30S-51S.

Gilson, S.C. (1989), Management Turnover and Financial Distress, Journal of FinancialEconomics 25, 241-262.

16

Greene, W.H. (2000), Econometric Analysis, 4th ed., New York.

Hall, B.H. (1992), Investment and Research and Development at the Firm Level: Does theSource of Financing Matter?, NBER Working Paper No. 4096, Cambridge.

Hall, B.H. (2002), The Financing of Research and Development, Oxford Review of EconomicPolicy 18(1), 35-51.

Hall, B.J. and J.B. Liebman (1998), Are CEOs Really Paid Like Bureaucrats?, QuarterlyJournal of Economics 113, 653-691.

Harris, M and A. Raviv (1990), Capital Structure and the Informational Role of Debt, Journal ofFinance 45, 425-458.

Hart, O. (2001), Financial Contracting, Journal of Economic Literature XXXIX(4), 1079-1100.

Hart, O. and J. Moore (1995), Debt and Seniority: An Analysis of the Role of Hard Claims inConstraining Management, American Economic Review 85, 567-585.

Henderson, R. (1993), Underinvestment and Incompetence as Responses to Radical Innovation;Evidence from the Photolithographic Alignment Equipment Industry, RAND Journal ofEconomics 24, 248-270.

Holmström, B. (1989), Agency Costs and Innovation, Journal of Economic Behavior andOrganization 12, 305-327

Holmström, B. and P. Milgrom (1987), Aggregation and Linearity in the Provision ofIntertemporal Incentives, Econometrica 55, 303-328.

Holmström, B. and P. Milgrom (1991), Multitask Principal-Agent Analyses: IncentiveContracts, Asset Ownership and Job Design, Journal of Law, Economics & Organization 7(Special Issue), 24-52.

Horn, H., H. Lang and S. Lundgren (1994), Competition, Long Run Contracts and InternalInefficiencies in Firms, European Economic Review 38, 213-233.

Jensen, M.C. (1986), Agency Costs of Free Cash Flow, Corporate Finance, and Takeovers,American Economic Review 76, 323-329.

Jensen, M.C. and K.J. Murphy (1990a), Performance Pay and Top-Management Incentives,Journal of Political Economy 98, 225-264.

Kaplan, S.N. (1994a), Top Executives, Turnover, and Firm Performance in Germany, Journal ofLaw, Economics & Organization 10, 142-159.

Kaplan, S.N. (1994b), Top Executive Rewards and Firm Performance: A Comparison of Japanand the United States, Journal of Political Economy 102, 510-546.

Kaplan, S.N. and P. Stromberg (2003), Financial Contracting Theory Meets the Real World: AnEmpirical Analysis of Venture Capital Contracts, Review of Economic Studies 70, 281-315.

Kirzner, I.M. (1985), Discovery and the Capitalist Process, University of Chicago Press:Chicago.

17

Kraft, K. and A. Niederprüm (1999), Determinants of Management Compensation with Risk-Averse Agents and Dispersed Ownership of the Firm, Journal of Economic Behavior andOrganization 40, 17-27.

Laffont, J.J. and J. Tirole (1986), Using Cost Observation to Regulate Firms, Journal ofPolitical Economy 94, 614-641.

Laffont, J.J. and J. Tirole (1990), The Regulation of Multiproduct Firms, Journal of PublicEconomics 43, 1-36.

Laffont, J.J. and J. Tirole (1993), A Theory of Incentives in Procurement and Regulation, MITPress, Cambridge, Mass.

Main, B.G.M. (1991), Top Executive Pay and Performance, Managerial and DecisionEconomics 12, 219-229.

Mehran, H. (1995), Executive Compensation Structure, Ownership, and Firm Performance,Journal of Financial Economics 38, 163-184.

Mises, L. (1951), Profit and Loss, Consumers-Producers Economic Services, South Holland.

Palley, T.I. (1997), Managerial Turnover and the Theory of Short-Termism, Journal ofEconomic Behavior & Organization 32, 547-557.

Perri, T.J. (1994), Influence Activity and Executive Compensation, Journal of EconomicBehavior & Organization 24, 169-181.

Ross, S.A. (1977), The Determination of Financial Structure: the Incentive-SignallingApproach, Bell Journal of Economics 8(1), 23-40.

Stulz, R. (1990), Managerial Discretion and Optimal Financing Policies, Journal of FinancialEconomics 32, 263-292.

Warner, J.B., R.C. Watts and K.H. Wruck (1988), Stock Prices and Top Management Changes,Journal of Financial Economics 20, 461-492.

Weisbach, M.S. (1988), Outside Directors and CEO Turnover, Journal of Financial Economics20, 431-460.

Zwiebel, J. (1995), Corporate Conservatism and Relative Compensation, Journal of PoliticalEconomy 103, 1-25.

Zwiebel, J. (1996), Dynamic Capital Structure under Managerial Entrenchment, AmericanEconomic Review 86, 1197-1215.